Instance-optimality in differential privacy via approximate inverse sensitivity mechanisms

Authors: Hilal Asi, John C. Duchi

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We study and provide instance-optimal algorithms in differential privacy by extending and approximating the inverse sensitivity mechanism. We provide two approximation frameworks... Finally, we use our approximation framework to develop private mechanisms for unbounded-range mean estimation, principal component analysis, and linear regression. The utility improvements in these examples demonstrate the advantages of our mechanisms over standard frameworks and the importance of these notions of instance-optimality.
Researcher Affiliation Academia Hilal Asi Stanford University asi@stanford.edu John C. Duchi Stanford University jduchi@stanford.edu
Pseudocode Yes Algorithm 1: Sampling from approximate inverse sensitivity; Algorithm 2: Private PCA using approximate inverse sensitivity; Algorithm 3: Gradient mechanism for linear regression (and Algorithm 4 in Appendix D.1).
Open Source Code No The paper does not contain any explicit statement about making source code available or provide any links to a code repository.
Open Datasets No The paper discusses applications to problems like mean estimation ("Given xi iid P with unbounded range") and linear regression ("we have data points (xi, yi) Rd R"), but it does not specify or provide access information for any public or open datasets.
Dataset Splits No The paper does not provide specific dataset split information (e.g., exact percentages, sample counts, or a detailed splitting methodology for training, validation, and test sets).
Hardware Specification No The paper does not provide any specific hardware details (e.g., exact GPU/CPU models or memory specifications) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details, such as library names with version numbers, needed to replicate the experiments.
Experiment Setup No The paper primarily focuses on theoretical analysis and algorithm design rather than empirical implementation details. It does not provide specific hyperparameter values, training configurations, or system-level settings for its applications.